v1.82.0-rc.1 is a major release candidate that contains major efficiency improvements for your PromQL backends, new ETL-persistent storage options, and new cost allocation features! To pick up a release candidate, add "--devel" to your installation instructions, eg: helm install kubecost kubecost/cost-analyzer --namespace kubecost --devel
rc.1 fixes a few javascript regressions from rc.0
Hourly ETL:
- Our cache is now progressively built on hourly data.
This should reduce the load on your PromQL backend and give better granularity for data. After 48h by default, data in the ETL is merged into daily windows - S3Store (beta):
Our ETL cache can now be backed to s3 instead of a persistent volume or in-memory, to better support users who do not want to run PVs (docs: https://github.com/kubecost/docs/blob/master/etl-s3-backup.md)
Support proportional allocation of SharedTenancyCosts: costs which cannot be associated with a single allocation but are part of the cluster or node that an allocation is a tenant of, eg cluster management costs and node attached disks - kube-state-metrics:v2.0+ support (beta)
While we are not yet shipping kube-state-metrics:v2.0+ with kubecost itself, we now support users who are emitting kube-state-metrics:v2.0+ to a bring-your-own promQL endpoint. Note that some deprecated metrics have moved to being emitted in kubecost itself— these will roll off over time. - Fix issue with some alert fields not being updated via configmap
- Fix issue where volume ids were not captured for AWS CSI disk provisioners
- Fix issue where some Azure billing data CSVs could not be parsed
- Add different alert aggregations (Issue: kubecost/cost-model#737 relevant docs PR: https://github.com/kubecost/docs/pull/93/files)